Construction of context-adjusted pretty output.

This package provides a set of data structures, classes and operators that facilitate the construction of a Prettyprinter Doc object. The difference between this an Prettyprinter is:

• Additional "saytag" parameter that can be used to control the Doc rendering.

• Brevity of syntax (using operators) designed to enhance conversion of arguments and readability of messages and conversion

This module provides a set of data structures, classes, and operators that facilitate the construction of a Prettyprinter Doc object.

Motivation

Standard prettyprinting is a monotonic conversion that does not allow for customization for different uses or environments. For example, when debugging, full and explicit information about a structure should be generated, but for checkpoint logging, a simple overview is usually more appropriate.

This library provides for an additional type parameter that can be used to control the conversion to a suitably verbose Prettyprinter Doc representation.

This is also highly useful in conjunction with logging to generate successively more verbose information as the logging verbosity increases.

Usage

Typical usage is to create a sayable message using the operators defined here and then extract Prettyprinter Doc from the saying and convert it to a printable format (here, simply using show for the default Prettyprinter rendering).

import qualified Prettyprinter as PP

foo :: Members '[ Logging SayMessage, Config ] r -> a -> b -> Eff r [b]
foo arg1 arg2 =
   do putStrLn $ show $ saying $ sayable @info "Entering foo with" &- arg1 &- "and" &- arg2
      rslt <- something arg1 arg2
      case rslt of
        Right vals ->
          do putStrLn $ show $ saying $ sayable @"verbose"
                 $ "Foo successfully returning" &% length vals &- "results:" &- vals
             return vals
        Left err ->
          do putStrLn $ show $ saying $ sayable @"error"
                 $ "Foo error (" &- arg1 &- PP.comma &- arg2 &- ") is" &- err
             throwError err

There are three messages printed: one on entry and one on either the success or failure paths. Each message may have different levels of information reported for the various arguments.

The saytag type parameter

Each sayable message uses a TypeApplication to specify a saytag which should be used for controlling the rendering of that message. This parameter is polykinded to provide maximum flexibility, but the most common kind is Symbol (e.g. "info", "verbose", "error", etc.).

Another frequent kind used for the saytag is GHC.TypeNats.Nat, allowing for an ordering of saytag types. However be aware that any instance constraints (e.g. saytag <= 9) are only resolved after the instance head is matched, so if the constraints do not match no other instances will be tried an an error is generated. Thus, rather than use constraints for selecting between instances, the maximum value for each "range" should be an instance, along with the minimum extremum:

instance {-# OVERLAPPING #-} Sayable (9::Nat) Foo where sayable f = ...[sayable for 9+]
instance {-# OVERLAPPING #-} Sayable (3::Nat) Foo where sayable f = ...[sayable for 3-8]
instance {-# OVERLAPPING #-} Sayable (0::Nat) Foo where sayable f = ...[sayable for 0-2]
instance {-# OVERLAPPABLE #-} (0 <= prevVer, prevVer ~ (ver - 1), Sayable prevVer Foo)
   => Sayable ver Foo where
   sayable = Saying . saying . sayable @Nat @prevVer

As a developer, it is encouraged to use whatever saytag makes sense relative to the current context and type of information being processed. Most of this documentation will use the preferred Symbol kind for the saytag.

Individual Arguments

The arguments passed to the sayable should be instances of the Sayable class. There are a number of standard instances of Sayable, but an instance can be declared for any object that might be output. The Sayable class has two class parameters: the second is object to be converted, and the first is the "saytag". This allows different Sayable instances for an object to be used in different saytag scenarios. For example:

import Network.URL

instance Sayable "verbose" URL where
  sayable url =
    let newline = PP.line :: PP.Doc SayableAnn
        prettyShow x = PP.viaShow x :: PP.Doc SayableAnn
    in "URL {"
        &- "url_type=" &- prettyShow (url_type url) &- newline
        &- "url_path=" &- url_path url &- newline
        &- "url_params=" &* url_params url
        &- "}"
instance Sayable saytag URL where
  sayable = Sayable . PP.viaShow . exportURL

The above would cause a url emitted via a "verbose" saytag to be expanded into a report on each individual field, whereas all other saytags would simply output the exportURL representation of the URL.

>>> let host = Host (HTTP True) "github.com" Nothing
>>> url' = URL (Absolute host) "by/one"
>>> saying $ sayable @"verbose" url'
URL { url_type= Absolute (Host {protocol = HTTP True, host= "github.com", port= Nothing})
 url_path= by/one
 url_params= }
>>> saying @"info" $ sayable url'
https://github.com:442/by/one

Note that there are several pre-declared Sayable instances for common datatypes for convenience.

Operators

In the logging lines above, there are several operators used, each of which starts with the & character. These are described in detail in the 'Helper operators' section below, but the general mnemonic for these is:

• A dash is a space between sayable elements

• A plus is immediately adjacent sayable elements

• A colon is a separator specification

• An asterisk is applied to a foldable (i.e. a list)

• A percent sign preceeds a Pretty object

• An exclamation follows a Pretty function, which is applied to the following argument.

• A question mark is followed by a Maybe, with no output for a Nothing

• A less-than character means newline (i.e. return to the left)

These characters will be combined for operators with combination effects.